Appendix: Further details on method

 

Response rate

The overall response rate was 26.8% (n=1,001). During fieldwork, an interviewer error in the UK resulted in the CATI (telephone methodology software) link being sent to web (self-administered) respondents. After reviewing the data, N=403 cases were removed from the data due to this error. If the correct link had been sent, the response rate would have been 37.6%.

Weighting

For the 11 country results, data from each country were weighted to ensure the final outcome was representative of GPs in that country based on their demographics (gender, age – and region for the UK) and selected specialty types.

This procedure also accounted for the sample design and probability of selection as needed, as well as differential non-response across known population parameters.

To create all England totals, the total number of respondents weighted for London and England (excluding London) were used to ensure the final outcome was representative of the number of London GPs versus rest of England GPs.

When creating an all respondent total (all 11 countries) or all other 10 countries total (to show in comparison to the UK result), the mean average percentage of the 10 or 11 percentages was calculated.

Uncertainty and limitations

Many of the survey results provide comparisons between countries. The weight-adjusted margin of error (for a 95% confidence level) varies between the 11 countries from 2.0% (Sweden) to 4.6% (Australia). For the UK as a whole it is 3.5%. This includes the ‘design effect’ (error introduced due to the weighting procedure).

This report includes a small number of sub-analyses between the different countries of the UK, and within different subgroups of respondents. Not all the results of these sub-analyses are significant at the 95% level, but we have included them as they may be of interest as part of a wider trend, and stated when they are not significant.

The margin of error (for a 95% confidence level) varies significantly between the different countries of the UK and it is important to bear this in mind when interpreting the findings. The margins of error for the five UK regions are:

  • England excluding London (475 respondents): 4.5%
  • London (200): 7.0%
  • Scotland (135): 8.5%
  • Wales (111): 9.3%
  • Northern Ireland (80): 11.0%

These are the theoretical margins of error if the percentage of respondents giving a certain answer is exactly 50%, where margins of error will be highest. They therefore give some indication of where particular caution should be taken with results. A margin of error is a relationship between sample size and the percentage of respondents giving a certain answer: it does not take into account how the survey was conducted.

Sampling error is only one type of error that affects survey outcomes. Other forms of error that are common to all surveys include selection bias (due to the respondents not being representative, and the weighting failing to fully address this) and questionnaire effects (such as questions being unclear or understood differently).


** The margin of error is one side of a confidence interval. So a margin of error of 3.5% for a 95% confidence level means that if the survey were conducted 100 times, you would expect the data to be within 3.5 percentage points above or below the percentage reported in 95 of the 100 surveys.

Previous